In the process of remote sensing image acquisition, low quality and lack of important information of image are common problems as the existence of interference information. Traditional image enhancement methods often cannot highlight useful information with high precision and high efficiency because they cannot integrate global information effectively. In order to solve these problems, a remote-sensing image enhancement method based on tensor decomposition and nonsubsampled Contourlet transform is proposed. The optimized nonsubsampled Contourlet transform is used to decompose the original image, and the high-order tensor is composed of high-frequency detail images in all directions on all scales. Through Bayesian probability tensor completion, the potential factors recognized from the incomplete tensor are used to predict the missing details of the image. Experimental results indicate that the proposed method can recover the missing information more effectively and highlight the feature information of the image. Compared with different image enhancement methods, the maximum improvement of signal-to-noise ratio, structure similarity and root mean square error are 27.9%, 37.6% and 45.4%, respectively. The proposed method is superior to the common image enhancement methods in quantitative evaluation and visual comparison.